ChatGPT 4.5 is out, and the internet shat all over it. But I think Sama might be onto something—it’s the first model that would fool me in a Turing test encounter.

Its less robotic nature, combined with gentle nuances and a fewer bullet-point answers, gives it a kind of emotional consciousness that no previous model has done for me.

Of course, hype-hungry benchmark evaluators hate this kind of thing because gentle subjectivity doesn’t translate well into leaderboard points.

But not everything important is quantifiable, and taste matters to some people more than others.